The Role of Information Technology in Health Literacy Research

Abstract

Without concerted effort, the current explosion in health information technology will further widen the digital health divide for individuals with inadequate health literacy. However, with focused investment of time and energy, technology has the potential for reducing disparities through intelligent, usable, and accessible systems that tailor information, advice, counseling, and behavioral support to an individual's need at a given time and place.

The development of information technology (IT) for patient and consumer health applications has been exploding in the past decade, with thousands of websites, hundreds of mobile applications, and dozens of special-purpose devices targeted at the health care market. However, this growth in technology is actually likely to increase health disparities for those with limited health, computer, and reading literacy, unless effort is devoted to the development of IT specifically designed for these disadvantaged groups, and issues of technology access are addressed. The design of technology for low-literacy individuals needs to be primarily focused on improved interface design, but all aspects of usability of health information systems must be addressed, including which health messages are delivered at a given time. There is now ample evidence that individuals with low health literacy have difficulty accessing and using state-of-the-art digital health communication media, such as websites (Jensen, King, Davis, & Guntzviller, 2010; Neter & Brainin, 2012; Neuhauser & Kreps, 2008; Sarkar et al., 2010). Barriers to health IT access, including cost, high-speed Internet access, access to consumer and patient medical devices, and rural-urban and regional variability in access, are often ignored by researchers developing new health technologies, but their resolution is crucial to ensure the ultimate success of their efforts.

If these issues of usability and access are addressed, IT has the potential to reduce health literacy-based disparities through the use of intelligent but easy-to-use systems that can inform, teach, persuade, and counsel, using messages that are perfectly tailored for an individual to understand and act upon in a given situation. There is now a growing body of research into the development and evaluation of computational artifacts designed specifically for populations with low health literacy that demonstrates the potential of IT to level the playing field by providing information at the place and time it is needed and in a format that is more understandable and actionable than traditional print media. Several of these artifacts have even been shown to be preferred over oral communication with health providers because they have the potential to provide a low-pressure context in which patients—especially those with low health literacy—do not feel talked down to and are open to asking questions (Bickmore, Pfeifer, & Jack, 2009).

In the rest of this article, we briefly survey the state of the art in using technology to help individuals with inadequate health literacy: automated assessment of individual health literacy level and document readability, and automated interventions for providing health information and/or counseling to individuals with inadequate health literacy. We close with some thoughts on future opportunities for the use of technology to assist individuals with low health literacy.

Information Technology in Health Literacy Assessment of Individuals

Several researchers have developed IT systems for automatically assessing the health literacy of individuals. Lobach and colleagues (Lobach, Arbanas, Mishra, Campbell, & Wildemuth, 2004; Lobach, Hasselblad, & Wildemuth, 2003) developed a computer-administered questionnaire to automatically assess users' reading and computer literacy levels, then used this assessment to automatically tailor the user interface to these factors. Yost and colleagues (2009) also developed a talking touchscreen system that assessed health literacy in English and Spanish, with the objective of completing assessments more efficiently and more accurately than the standard Test of Functional Health Literacy in Adults. To date, these efforts have been designed as research tools. In the near future, computer-adaptive testing will shorten the duration of these assessments and make them even easier to use. This could potentially lead to clinical use if and when a given patient's health literacy status needs to be identified in order to trigger specific interventions, although this has not yet occurred as of the time of writing.

Information Technology in Readability Assessment

Other researchers have developed automated approaches to assessing the readability levels of printed documents and websites. Although the Flesch-Kincaid readability calculator is the most well known given that it is a feature in Microsoft products, Dale-Chall, Gunning-Fog, SMOG, and other algorithms for automatically assessing the readability of documents are available in most modern word processors (Ley & Florio, 1996). These standard measures are easy to compute and use but are not reliable, as demonstrated by several recent studies (e.g., Feng, Jansche, Huenerfauth, & Elhadad, 2010). To date, they have been good tools to identify blocks of text that are likely to be inadequate. However, a transformation may be taking place as more sophisticated algorithms, on the basis of techniques from computational linguistics have yielded more accurate results (Feng et al., 2010). Preliminary research has also been conducted into automatically mapping lay (“consumer”) health care terms to standardized medical nomenclature (Zeng & Tse, 2006). This work lays the groundwork for systems that can eventually be developed to automatically translate between technical language and language that people can better understand.

Information Technology–Based Interventions

Several computerized health interventions have now been developed and evaluated specifically for individuals with low health literacy, such as multimedia touch screen computers that administer standard diagnostic questionnaires (Bryant et al., 2009; Wofford, Currin, Michielutte, & Wofford, 2001), and health education systems that are designed to accommodate a wide range of health, computer, and reading literacy levels (Gerber et al., 2005; Kandula et al., 2009). These systems commonly provide information using audio, video or graphical displays along with text to minimize the literacy level required of users, and often use touch screen input so that users do not need to have mouse and keyboard proficiency (Gerber et al., 2005; Kandula et al., 2009). For example, the Multimedia Diabetes Education Program uses images, 2D animation, and spoken audio (narrator) along with text. In a study of 190 patients (21% with low health literacy), it was found that all patients had significant increases in diabetes knowledge after interacting with the program, although the program did not fully compensate for low health literacy.

Interactive voice response (or phone-based) systems represent another modality that may be particularly well-suited to deliver interventions to low-literacy individuals because these systems (a) deliver information in speech instead of text and (b) use the ubiquitous communication channel of the telephone so that participants neither need computer literacy nor computer access (Piette, 2000). Schillinger and colleagues (2008) compared a multilingual interactive voice response system for diabetes self-management support with monthly group medical visits, and demonstrated that the interactive voice response system resulted in greater engagement, especially among participants with limited literacy. Shea and colleagues (2008) performed a direct comparison of a health satisfaction survey (Consumer Assessment of Healthcare Providers and Systems) delivered either in its original 63-item text format, the text format augmented with images, or a bilingual interactive voice response–based implementation, and demonstrated that significantly more individuals with inadequate health literacy completed the survey, and completed it is less time and with fewer errors, when delivered through interactive voice response compared with the other two modes.

Our own research in this area involves the use of embodied conversational agents that appear on the computer as animated characters that simulate the conversational nonverbal behavior that accompanies human face-to-face dialogue, such as hand gestures, facial displays, and gaze (Figure 1). We have found that patients learn more with embodied conversational agents compared to other modalities (e.g., human instructors), regardless of health literacy, and that individuals with inadequate health literacy generally report higher levels of satisfaction with embodied conversational agents and ask more questions with embodied conversational agents compared to individuals with adequate health literacy (Bickmore and Paasche-Orlow, 2011). However, as with the Multimedia Diabetes Education Program, embodied conversational agents did not fully compensate for low health literacy (Bickmore, Pfeifer, & Paasche-Orlow, 2009; Bickmore et al., 2010).

Figure 1 Example of embodied conversational agent used in walking promotion intervention for older adults (Bickmore et al., 2010). (Color figure available online.).

Future Directions for Information Technology in Health Literacy

There are a great many promising directions of research into how technology can improve the health of individuals with inadequate health literacy. As described earlier, most of the work to date in this area has been on the development of technologies designed to improve an individual's knowledge about specific areas of health and self-care, such as diabetes. Even in this area, there is a tremendous opportunity for new research into improved pedagogical methods that can be borrowed from the field of Intelligent Tutoring Systems and applied in specific health domains (Nkambou, Bourdeau, & Mizoguchi, 2010). In addition to improving knowledge, intelligent systems can be used to improve skill deficits in health care (e.g., inhaler technique) as well as underlying deficits in basic literacy skills. For example, automated reading tutors that listen to learners read out loud and provide coaching (Mostow et al., 2003) and math tutors that provide adaptive, personalized instruction to improve numeracy skills have both been developed and tested (Melis & Siekmann, 2004). IT systems can also play a role in providing access to health care and other services. For example, intelligent online directories have been designed to help connect urban, low income populations with community services including health care, housing, food security, and income security services (Fleegler, Lieu, Wise, & Muret-Wagstaff, 2007). Extending this kind of matchmaking service to add educational support for people who are struggling with their care is an important research direction.

Intelligent counseling systems can go beyond the provision of knowledge and literacy skills to impact a wide range of other psychological constructs associated with health literacy. Just as a good human counselor can work with an individual to improve self-efficacy and motivation to change a health behavior, automated counseling agents can use techniques from health behavior change to improve health behaviors over time, and the messages and techniques used could be tailored to the individual's literacy (Bickmore & Giorgino, 2006; Bickmore, Schulman, & Sidner, 2011). Counseling agents could also be used to change norms by modeling or demonstrating ideal health behavior (e.g., medication adherence, walking, weight loss), and help prevent the discounting of future effects of current health decisions by reminding individuals of the long-term ramifications of their actions. Agents could also help address the fatalism about health outcomes many individuals with inadequate health literacy embrace by making individuals aware of the choices they have in their care and changing their expectations about the future.

Information systems could be used to significantly improve patient–provider communication for individuals with low health literacy. Interventions to promote patient activation are particularly important for low-literacy patients because they tend to ask fewer questions in comparison with individuals with adequate literacy (Katz, Jacobson, Veledar, & Kripalani, 2007; Schillinger et al., 2003). Mobile systems that maintain lists of questions and issues for patients and promote their resolution during medical encounters represent one of many ways that IT systems could become automated advocates for those who lack family or friends who can perform this role. Automated coaches could also give patients practice interacting with virtual health providers to build self-efficacy for interactions with the actual health care system. Automated systems could also be used to help facilitate and mediate patient–provider interactions by, for example, negotiating the agenda of problems to address with providers and by facilitating shared decision making. Automated systems could also be used to change the behavior and attitudes of providers themselves when interacting with low-literacy patients by, for example, prompting them to evaluate comprehension and other best practice interventions.

Last, IT systems can be used to significantly enhance the social support networks for individuals with low health literacy. Existing websites such as PatientsLikeMe.com and the many condition-specific support websites are wonderful resources for those who are connected into the digital world, but these resources are largely out of reach for those with low computer literacy and limited resources. In addition, the members of these online communities tend be well educated and highly health literate, which may represent a barrier for low-literacy individuals to fully participate. Parallel resources need to be established for those who lack this digital and health fluency and lack access to IT systems to connect them both to expert health advice and peers who can provide emotional, informational, and instrumental support.

Conclusion

Imagine a future in which an individual can carry a mobile device (perhaps their smartphone) that not only provides health information, but senses and interprets the environment and makes proactive recommendations about actions that might affect their health. The device could also translate health-related text into actionable messages, tailored not only to the individual's health literacy level, but their educational and cultural background, values, and even emotional state, providing information and recommendations at the time and location, and in the format, that will have the most effect. Such a digital health conscience would need a detailed cognitive model of its wearer, going beyond canonical models of health literacy and culture—which can simply reinforce stereotypes—to a detailed representation of a user's unique desires, beliefs, and behavior patterns, and how these have changed over time. Elements of such systems already exist, and limited versions are certainly in our near future. However, the market for health IT products will not take care of our most vulnerable populations on its own; we need continued, focused, evidence-based research to continue developing technologies that specifically address disparities for individuals with low health literacy in order to realize this potential.